Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Fruit 1.5 parameter test extended

Author: Volker Böhm

Date: 01:14:30 08/28/04

Go up one level in this thread


Hi Stuart,

as everytime there is no "right" or "false" for nullmove condition. It depends
on your chess engine. I have the following conditions, the nullmove is not done
if:

1. Nullmove has been done right before
2. hash entry says the value is < beta (I use even hash entries if
remaining_depth <= hash_remaining_depth + 3) to thest this condition - it didn´t
work for a long time, hazardous it works now - This is a "poor-man´s" verified
nullmove :-)
3. Remaining Depth <= 1
4. Material + Pawn < Beta (eval + Pawn < Beta works less good for spike - seems
material is a more stable condition than positional value if not moving
5. King is in Check
6. Beta <= MaxVal - Ply (Beta says player to move is forced mate in the next
ply)
7. If player to move can mate in one move (have a static test that finds ca. 95%
conditions)
8. King has been in Check in ply - 2 and ply - 4 (idea from gnu-chess I don´t
know if it is worth something - not enough tests)
9. Not if player to move has no range piece (queen, rook, bishop)
10. Not if root-search-depth is below 3

Greetings Volker




This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.